What problem does it solve? Workspace guidance files, skills, and memory indexes often accumulate outdated context-engineering patterns like redundant rules, duplicated instructions, and bloated always-loaded files, degrading how effectively Claude Code operates. ## Core Features & Use Cases - Standard Health Check: Runs the built-in claude doctor diagnostic with a 60-second timeout and summarizes findings without blocking on interactive screens. - Six-Shift Context Audit: Evaluates guidance files, skills, rules, and memory indexes against Anthropic's six context-engineering shifts: judgement over rules, interfaces over examples, progressive disclosure, one home per instruction, auto-memory, and rich references. - Approval-Gated Fixes: Reports a findings table with verdicts, worst offenders, and suggested fixes first, applying changes only after explicit user approval. - Use Case: After months of accumulating CLAUDE.md instructions and custom skills, run a context checkup to find duplicated guidance, over-constraining rules, and facts that belong in auto-memory instead of guidance files. ## Quick Start Ask the agent to run a context checkup on this workspace and audit my guidance files against the six context-engineering shifts.
